A new dental panoramic X-ray image registration technique using hybrid and hierarchical strategies

In this paper, a wavelet-based decomposition level with scale invariant feature transform (SIFT) is proposed to solve the problem of image registration. At the highest wavelet-based pyramid level, mutual-information (MI) based registration is applied and rough similarity (linear conformai) transformation parameters are achieved. At the first wavelet-based decomposition level, SIFT-based registration technique is utilized with the aid of rough parameters obtained. To remove outliers automatically a RANdom Sample And Consensus (RANSAC) algorithm is applied. A comparison between proposed technique with three registration approaches is achieved: cross-correlation based registration, point mapping image registration, and hybrid registration technique using MI and SIFT in the spatial domain. The quality of the registration process was measured using the following criteria: normalized cross-correlation coefficient (NCCC) and percentage relative root mean square error (PRRMSE). The application of the proposed technique to dental panoramic X-ray images has shown that wavelet-based hybrid approach combining MI and SIFT operator achieves high performance registration results and can be used efficiently for registration of X-ray images.